Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/10835#discussion_r50805180
--- Diff: core/src/main/scala/org/apache/spark/executor/TaskMetrics.scala
---
@@ -230,86 +297,119 @@ class TaskMetrics extends Serializable {
*/
def shuffleWriteMetrics: Option[ShuffleWriteMetrics] =
_shuffleWriteMetrics
- @deprecated("setting ShuffleWriteMetrics is for internal use only",
"2.0.0")
- def shuffleWriteMetrics_=(swm: Option[ShuffleWriteMetrics]): Unit = {
- _shuffleWriteMetrics = swm
- }
-
/**
* Get or create a new [[ShuffleWriteMetrics]] associated with this task.
*/
private[spark] def registerShuffleWriteMetrics(): ShuffleWriteMetrics =
synchronized {
_shuffleWriteMetrics.getOrElse {
- val metrics = new ShuffleWriteMetrics
+ val metrics = new ShuffleWriteMetrics(initialAccumsMap)
_shuffleWriteMetrics = Some(metrics)
metrics
}
}
- private var _updatedBlockStatuses: Seq[(BlockId, BlockStatus)] =
- Seq.empty[(BlockId, BlockStatus)]
-
- /**
- * Storage statuses of any blocks that have been updated as a result of
this task.
- */
- def updatedBlockStatuses: Seq[(BlockId, BlockStatus)] =
_updatedBlockStatuses
- @deprecated("setting updated blocks is for internal use only", "2.0.0")
- def updatedBlocks_=(ub: Option[Seq[(BlockId, BlockStatus)]]): Unit = {
- _updatedBlockStatuses = ub.getOrElse(Seq.empty[(BlockId, BlockStatus)])
- }
+ /* ========================== *
+ | OTHER THINGS |
+ * ========================== */
- private[spark] def incUpdatedBlockStatuses(v: Seq[(BlockId,
BlockStatus)]): Unit = {
- _updatedBlockStatuses ++= v
+ private[spark] def registerAccumulator(a: Accumulable[_, _]): Unit = {
+ accums += a
}
- private[spark] def setUpdatedBlockStatuses(v: Seq[(BlockId,
BlockStatus)]): Unit = {
- _updatedBlockStatuses = v
+ /**
+ * Return the latest updates of accumulators in this task.
+ */
+ def accumulatorUpdates(): Seq[AccumulableInfo] = accums.map { a =>
+ new AccumulableInfo(
+ a.id, a.name.orNull, Some(a.localValue), None, a.isInternal,
a.countFailedValues)
}
- @deprecated("use updatedBlockStatuses instead", "2.0.0")
- def updatedBlocks: Option[Seq[(BlockId, BlockStatus)]] = {
- if (_updatedBlockStatuses.nonEmpty) Some(_updatedBlockStatuses) else
None
+ // If we are reconstructing this TaskMetrics on the driver, some metrics
may already be set.
+ // If so, initialize all relevant metrics classes so listeners can
access them downstream.
+ {
+ var (hasShuffleRead, hasShuffleWrite, hasInput, hasOutput) = (false,
false, false, false)
+ initialAccums
+ .filter { a => a.localValue != a.zero }
+ .foreach { a =>
+ a.name.get match {
--- End diff --
Underindented?
If this was done for line-length reasons, you could add a
`.map(_.name.get)` after the `.filter` then do `.foreach { case` and use the
implicit match statement to avoid having to add more indentation.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
